Introduction

Interleukin (IL)-1β initiates inflammatory reactions and amplifies immunologic responses against harmful stimuli [1]. Furthermore, in chronic inflammatory states, IL-1β generates COX-2 and iNOS, which inhibit apoptosis, induce DNA damage, and modulate cell adhesion [2]. In addition, the signaling cascade from IL-1β is the basis of the carcinogenesis. In addition to persistent inflammatory reaction caused by gastric injury, IL-1β suppresses acid secretion 6,000 times as effectively as H2 antagonist and 100 times more than proton pump inhibitor [3]. The expression of this cytokine creates a hypoacidic condition that favors the survival of Helicobacter pylori (H. pylori), and consequently leads to atrophy of stomach tissues or adenocarcinoma [4, 5], and overgrowth of H. pylori induces an assembly of neutrophils and lymphocytes, particularly Th1 and Th17 CD4+ cells, which induce IL-1β secretion [6, 7]. Three single-nucleotide polymorphisms (SNPs), namely −31 T/C, +3954 C/T, and −511 C/T have been discovered in the promoter region of chromosome 2q and are regarded to trigger the overexpression of IL-1β [8]. This study focuses on the IL-1β-511 C/T polymorphism because many studies have investigated it, whereas relatively few have examined the +3954 C/T polymorphism. In addition, the −31 T/C SNP has been reported to show linkage disequilibrium with −511 C/T [9].

However, previous studies, including meta-analyses, have produced mixed results [10–15], which may have been caused by dissimilar characteristics among studies, such as, sample sizes, ethnicities, cancer type, inconsistent inclusion criteria (e.g., involving premalignant lesions as a case group), and a lack of comprehensive subgroup analyses. In this regard, the present study provides a comprehensive and systematic review based on sophisticated subgroup analysis that excluded methodological discrepancies. In addition, overall susceptibility results were verified by sensitivity analysis based on considerations of Hardy–Weinberg equilibrium (HWE) as a crucial standard for determining the reliability of subject for case–control studies [16].

Materials and methods

Search strategy

A systematic search was conducted utilizing the MEDLINE, EMBASE, and CENTRAL databases (last search on February 05, 2013). The following terms were combined: “Interleukins,” “IL-1β,” “IL-1β-511,” “Interleukin-1,” “Interleukin-1β,” “Interleukin-1β-511,” “Interleukin-1beta,” “IL-1beta,” “Interleukin-1beta,” “polymorphisms,” “SNP,” “single nucleotide,” “mutation,” “stomach cancer,” “gastric adenocarcinoma,” and “gastric cancer.” Supplement S1 describes the detailed search strategy, which was reviewed by two independent investigators (M.J.P and M.H.H) and a third reviewer (S.S.P).

Study selection

The studies included: (i) described the relationship between the IL-1β-511 C/T SNP and stomach carcinoma; (ii) contained sufficient number of subjects to yield odds ratios (ORs) and 95 % confidence intervals (CIs); (iii) had a case–control design; (iv) included case samples consisting of gastric cancer (not premalignant lesions), and control samples free of any gastric disease, such as, gastritis or gastric ulcer; and (v) were written in English. A PRISMA checklist and a flow chart of the study inclusion procedure are presented in Supplements S2 and S3, respectively.

Methodological quality assessment

The methodologic quality of each study was assessed using the scale proposed by Thakkinstian et al. [7] and refined by Camargo et al. [12] and Xue et al. [13]. Any disagreement between evaluation results was resolved by the third reviewer (S.S. Park). Evaluations were conducted to determine the representativeness of cases and controls, to assess reliability of stomach carcinoma confirmation and genotyping tests, and to assess potential confounding factors, as shown in Supplement S4. Quality assessment scores ranged from 0 (lowest) to 9 (highest). We classified reports that scored <5.0 as “low to moderate quality” and those that scored ≥5.0 as “high quality.”

Data extraction

To enhance the reliability of data, two investigators (M.J. Park and M.H. Hyun) independently performed and verified data extraction. The following information was collected: authors’ names, subject ethnicity, sex ratio, origin of control samples, numbers of cases and controls, and the genotyping method. In addition, the genotype frequencies of each pathologic type of cancer, each anatomical classification of cancer, and of H. pylori-positive populations were determined when reports provided relevant information.

Statistical analysis

We utilized Review Manager 5.2 (Cochrane Collaboration, London, UK) to conduct the statistical analysis. ORs and 95 % CIs were calculated from extracted raw data, and strengths of association were estimated [17]. Meta-analysis was conducted using the following models: (1) T allele versus C allele (an allelic contrast model), (2) TT genotype versus CC genotype (a homozygote contrast model), (3) TT+TC genotype versus. CC genotype (a dominant contrast model), and (4) TT vs. TC+CC (a recessive contrast model).

Heterogeneities of included studies were calculated based on Q statistics using the Mantel–Haunszel weight and I 2 statistics. [18]. Heterogeneity between studies was confirmed when studies have a P value of <0.10 and an I 2 value >50 %. For studies with heterogeneity, a random effects model was employed based on the DerSimonian–Laird method [19]. Otherwise, a fixed effects model was employed based on the Mantel–Haenszel method [18].

We conducted Chi square analysis to assess the control group fit with the Hardy–Weinberg equilibrium (HWE). The groups deviating from the HWE have a P value of <0.05 [20]. Begg’s test and Egger’s funnel plot asymmetry test were used to evaluate publication bias [21, 22].

Results

Literature search, characteristics of included studies, and publication bias

The overall flow of the searching procedure is shown in Fig. 1. First, a total of 824 studies were identified by systematic search after excluding duplicates. Screening of full texts for relevance and accessibility resulted in the exclusion of 482 irrelevant studies and 7 abstract-only articles. 290 of the remaining 335 studies were excluded for the following reasons: 93 for including premalignant lesions, not gastric cancer, 138 for including control populations with gastritis or dyspeptic disease, 44 for not having a case–control design, and 15 for being written in other than English. In addition, 2 studies included data from two different geographic areas [23, 24]. Persson et al. [25] recruited control subjects from two sources: hospitals and general population. Zhao et al. [26] considered three ethnic groups from the same area. In the present study, each geographic area, control source, and ethnicity were considered separate data sets. As a result, 45 studies (50 data sets) were included in this meta-analysis, reflecting 9,066 gastric cancer patients and 11,192 control subjects [8, 23–66]. Table 1 shows the characteristics of each study. Of the 45 studies, control groups deviated from the HWE in 10 [33, 34, 36, 39, 41, 42, 51, 59, 60, 63]. In addition, 14 studies involved Caucasian populations, 27 Asian populations, and 4 other ethnicities. Twenty-four studies employed the PCR–RFLP genotyping, and the remainder used other genotyping techniques, such as, RCP–SSCP and PCR–DHPLC. Twenty-nine studies were classified as high quality, and 16 as low to moderate quality. Supplement S5 summarizes quality assessment criteria. Finally, we used a funnel plot and Egger’s regression to assess the heterogeneity of studies and publication bias. Figure 2 presents the qualitative results for publication bias and shows a symmetrical distribution for the overall studies. Egger’s regression revealed no publication bias (P > 0.1).

Fig. 1
figure 1

Study flow chart

Full size image
Table 1 Characteristics of included studies
Full size table
Fig. 2
figure 2

A funnel plot of publication bias

Full size image

Overall results on the relationship between the IL-1β-511 C/T SNP and gastric cancer

Figure 3 summarizes the results of sensitivity analysis based on the HWE principle using the recessive model (TT vs. CC+CT). For overall studies, interleukin 1β-511 C/T SNP was found to be positively related to the risk of stomach carcinoma (OR = 1.15; 95 % CI 1.03–1.29). Studies satisfying the HWE supported this relationship with a similar odds ratio (OR = 1.15; 95 % CI 1.01–1.29), whereas those deviating from the HWE showed no association between the IL-1β-511 C/T SNP and the risk of stomach carcinoma (OR = 1.19; 95 % CI 0.84–1.69).

Fig. 3
figure 3

A forest plot of the stomach carcinoma risk of relevance to the interleukin-1β-511 C/T polymorphism (TT vs. CC+CT) based on the Hardy–Weinberg equilibrium by publication year. The areas of the squares indicate the relative weights of the specific studies. Bars represent 95 % confidence intervals, and “GC,” gastric cancer

Full size image

Comprehensive subgroup analysis for overall studies and HWE studies

Table 2 summarizes the outcomes of comprehensive subgroup analysis with respect to ethnicity, study quality, control sources, genotyping methods, anatomical locations of cancer, pathologies of cancer, and H. pylori infection. When stratified by ethnicity, a positive relationship was observed for Asian populations for both overall (OR = 1.14; 95 % CI 1.01–1.29) and HWE satisfying (OR = 1.16; 95 % CI 1.01–1.33) studies. However, no such association was observed for Caucasian populations for overall (OR = 1.15; 95 % CI 0.87–1.52) and HWE satisfying (OR = 1.10; 95 % CI 0.85–1.42) studies (Fig. 4). H. pylori-positivity group was related to the risk of stomach carcinoma in overall (OR = 1.70; 95 % CI 1.03–2.80) and HWE satisfying (OR = 2.04; 95 % CI 1.15–3.62) studies (Fig. 5). In addition, PCR–RFLP genotyping method is better at revealing susceptibility of IL-1β-511 T allele carrier to gastric cancer than other PCR methods, such as, PCR–DHPLC and PCR–SSCP for both overall (OR = 1.24; 95 % CI 1.05–1.47) and HWE satisfying (OR = 1.28; 95 % CI 1.05–1.56) studies (Fig. 6). In terms of study quality, high-quality studies showed a correlation between IL-1β-511 T carrier and risk of stomach carcinoma. (OR = 1.17; 95 % CI 1.00–1.36) (Fig. 7). Anatomical location of cancer does not affect to the relationship between IL-1β-511 C/T SNP and stomach cancer in either overall (cardia OR = 0.98; 95 % CI 0.46–2.08, noncardia OR = 1.25; 95 % CI 0.60–2.60) or HWE satisfying (cardia OR = 0.66; 95 % CI 0.35–1.24, noncardia OR = 0.91, 95 % CI 0.66–1.26) studies (Fig. 8).

Table 2 Stratification analyses of the interleukin-1β -511 C/T polymorphism (TT vs. CC+CT) for stomach carcinoma risk according to the Hardy–Weinberg equilibrium
Full size table
Fig. 4
figure 4

A forest plot of the stomach carcinoma risk of relevance to the interleukin-1β-511 C/T polymorphism (TT vs. CC+CT) by ethnicity subgroups based on the Hardy–Weinberg equilibrium. The areas of the squares indicate the relative weights of the specific studies. Bars represent 95 % confidence intervals, and “GC,” gastric cancer

Full size image
Fig. 5
figure 5

A forest plot of the stomach carcinoma risk of relevance to the interleukin-1β-511 C/T polymorphism (TT vs. CC+CT) in the H. pylori-positive subgroup based on the Hardy–Weinberg equilibrium. The areas of the squares indicate the relative weights of the specific studies Bars represent 95 % confidence intervals, and “GC,” gastric cancer

Full size image
Fig. 6
figure 6

A forest plot of the stomach carcinoma risk of relevance to the interleukin-1β-511 C/T polymorphism (TT vs. CC+CT) according to genotyping method based on the Hardy–Weinberg equilibrium. The areas of the squares indicate the relative weights of the specific studies. Bars represent 95 % confidence intervals. GC gastric cancer, PCR polymerase chain reaction, RFLP restriction fragment length polymorphism

Full size image
Fig. 7
figure 7

A forest plot of the stomach carcinoma risk of relevance to the interleukin-1β-511 C/T polymorphism (TT vs. CC+CT) according to the study quality subgroup based on the Hardy–Weinberg equilibrium. The areas of the squares indicate the relative weights of the specific studies. Bars represent 95 % confidence intervals, and “GC,” gastric cancer

Full size image
Fig. 8
figure 8

A forest plot of the stomach carcinoma risk of relevance to the interleukin-1β-511 C/T polymorphism (TT vs. CC+CT) according to histology subgroups based on the Hardy–Weinberg equilibrium. The areas of the squares indicate the relative weights of the specific studies. Bars represent 95 % confidence intervals, and “GC,” gastric cancer

Full size image

Discussion

Gastric cancer maintains its second place position amongst the causes of cancer-associated mortality, and therefore, researchers worldwide have concentrated on unearthing its etiology. H. pylori is a major causative agent in gastric cancer and produces oxidative radicals, which are harmful to DNA stability and stimulate the secretion of the proliferative factor gastrin [1, 67]. Interleukin 1β (IL-1β) amplifies this mechanism through hypochlorhydria, which is favorable to H. pylori [4]. Nevertheless, the fact that not every H. pylori carrier develops stomach carcinoma strongly suggests a relation between IL-1β polymorphisms and stomach carcinoma susceptibility. Many studies have investigated this potential relationship since El-Omar et al. [8] first described a positive correlation between the IL-1β-511 C/T SNP and risk of stomach carcinoma. However, such studies and even meta-analyses have produced mixed results. In this context, the present meta-analysis draws comprehensive analysis regarding the strength of the relationship between the IL-1β-511 C/T SNP and gastric cancer risk by in-depth analysis and the removal of presumed factors of heterogeneity from previous studies. A total of 45 recent studies with 50 population data sets were considered after eliminating selection bias from unrefined searches and systematically searching the MEDLINE, EMBASE, and CENTRAL databases. In addition, clear criteria for excluding and including studies were set. As a result, 12 new studies were added, and the homogeneity of control groups was elaborated by eliminating studies that included premalignant gastric patients as an eligible control group. In addition, HWE studies were analyzed because a deviation from the HWE implies that the study may exhibit selection bias or have suffered some erroneous event during genotyping [16, 20]. Consequently, consistency with the HWE is critical for guaranteeing the appropriateness of control subjects for a given case control study and for verifying the credibility of a genotyping procedure.

In this meta-analysis, we found that the IL-1β-511 C/T SNP confers susceptibility to stomach carcinoma, which is in accordance with the results of six of the seven previous meta-analyses. It is noteworthy that some of our subgroup analysis results are inconsistent with the findings of previous meta-analyses.

With respect to ethnicity, Asian populations were found to show a positive relation between the presence of the IL-1β-511 T allele and the risk of stomach carcinoma in overall and HWE satisfying studies, which is inconsistent with the findings of three previous meta-analyses [10, 12, 13] that found no such relation. In the present meta-analysis, ten recently published Asian studies were included and five Caucasian studies were excluded because healthy individuals and patients with premalignant lesions such as ulcer, MALToma, and gastritis were not differentiated in control populations. In this regard, the large size of Asian populations included in the present study and the process used for selecting control populations probably affected the results.

In subgroup analysis of H. pylori carriers, it was shown that H. pylori infection reinforces the relation between IL-1β-511 T allele with susceptibility to gastric cancer in both overall and HWE satisfying studies. It provides support for the mechanism that IL-1β contributes to chronic inflammation by producing hypochlorhydria. Close attention should be paid to the fact that East Asian populations have generally high H. pylori infection rate and that Asian exhibited the strongest relationship between the IL-1β-511 C/T SNP and gastric cancer in our study. As explained above, IL-1β-511 C/T SNP makes IL-1β to be overexpressed, which suppresses acid secretion and thus creates favorable conditions for H. pylori proliferation. These findings suggest that Asian populations, which are particularly vulnerable to H. pylori infection, would have a strong relationship between the IL-1β-511 C/T SNP and gastric cancer.

In addition, the relationship between stomach carcinoma and the IL-1β-511 C/T SNP was noteworthy in studies that employed PCR–RFLP genotyping methods. Of the various genotyping methods used, PCR–RFLP method was employed most frequently (25 of the 50 population data sets). The PCR–RFLP method was the first DNA-profiling technique which was used for genetic fingerprinting, evaluating the risk of genetic disorders, and analyzing samples from crime scenes. The large pool of cases analyzed in previous studies using PCR–RFLP supports the validity of the method. In addition, the sensitivity of PCR–RFLP has been verified in many studies. Accordingly, the reliability and validity of this method support the observed relation between IL-1β-511 T allele and risk of gastric cancer.

A subgroup analysis according to study quality scores showed a positive association between the IL-1β-511 T allele and the risk of stomach carcinoma. Quality analysis was carried out utilizing the scale proposed by Thakkinstian et al. [7] and refined by Camargo et al. [12] and Xue et al. [13]. High-score studies were elaborated by designating the source of control groups, confirming the representativeness of cases, using reliable methods to confirm the presence of stomach carcinoma and to conduct genotyping. Classifying studies by quality reduced study heterogeneity because it ensured the reliabilities of control sources and case populations.

In the present study, only recessive model (TT vs. CT+CC) results are presented, but recessive model results are consistent with those of other models. In the additive model (TT vs. CC), pooled ORs (95 % CI) were 1.20 (1.00–1.45) for overall studies and 1.24 (1.02–1.57) for HWE satisfying studies. In addition, the dominant model (TT+CT vs. TT) produced insignificant results for overall studies (OR = 1.10; 95 % CI 0.97–1.24), but the susceptibility of IL-1β-511 T carriers for HWE satisfying studies (OR = 1.14; 95 % CI 1.01–1.27) (data not shown). A similar pattern was observed for ethnicity analysis. Caucasian populations showed no relationship for any model: TT+CT versus CC (overall: OR = 1.13; 95 % CI 0.89–1.42; HWE: OR = 1.12; 95 % CI 0.87–1.44) and TT versus CC (overall: OR = 1.23; 95 % CI 0.86–1.75; HWE: OR = 1.18; 95 % CI 0.87–1.62), whereas Asian in HWE satisfying studies showed statistical significance in all models: TT+CT vs. CC (OR = 1.15; 95 % CI 1.00–1.34) and TT versus CC (OR = 1.26; 95 % CI 1.04–1.54).

This meta-analysis has several limitations that should be considered. Although we evaluated publication bias comprehensively using a funnel plot, Egger’s test, and Begg’s test, the tendency not to publish negative results may have produced this bias. In addition, future research should provide updated systematic analysis on the relationship between gastric cancer and haplotypes of the gene family cluster on chromosome 2q, IL-1β-31 C, IL-1β +3954, and IL1RN, because a polymorphism in one gene is often accompanied instability of a nearby gene.

In summary, the results of this refined and updated meta-analysis verify the relationship between the IL-1β-511 T allele carrier and stomach carcinoma susceptibility. It also confirms that Asian ethnicity strengthens this relationship. In addition, the coexistence of IL-1β-511 C/T SNP and H. pylori infection was found to increase susceptibility to stomach carcinoma. The most reliable genotyping technique appears to be PCR–RFLP, which suggests that it should be used to analyze the relationship between the IL-1β-511 C/T SNP and the risk of stomach carcinoma. Thorough screening of eligible studies and the adoption of a strict study selection procedure based on the elimination of selection bias for control groups may explain reported inconsistencies across ethnic subgroups.